Miyami WebSearch MCP


> Connect your LLM to the internet! Search the web and extract content from any webpage using the Model Context Protocol.

🌟 Features

- 🔍 Web Search - Search across Google, DuckDuckGo, Bing, Brave, Wikipedia
- 🧠 Deep Research - Multi-query parallel research with compiled reports
- 🌐 Site Crawl - Depth-limited crawling with Trafilatura extraction
- 🎬 YouTube Transcripts - Fetch captions/subtitles from any YouTube video - NEW!
- 🛡️ FREE Stealth Mode - Anti-bot bypass (Cloudflare, DataDome, etc.)
- ⏰ Time-Range Filters - Filter results by recency (day, week, month, year)
- 📄 Enhanced Content Extraction - Trafilatura-powered (Firecrawl-quality) extraction
- 📝 Markdown Output - Get structured markdown from webpages
- 🎯 Rich Metadata - Automatically extract authors, dates, site names
- ⚡ Fast & Easy - One-line installation, zero configuration
- 🤖 LLM Optimized - Formatted responses perfect for AI consumption
- 🆓 100% Free - No API keys, no signup, no configuration needed
- 🔒 Privacy-First - No tracking, no data collection

📦 Installation

$3

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

``json
{
"mcpServers": {
"miyami-websearch": {
"command": "npx",
"args": ["-y", "miyami-websearch-mcp"]
}
}
}
`

$3

`bash
npm install -g miyami-websearch-mcp
`

Then configure Claude Desktop:

`json
{
"mcpServers": {
"miyami-websearch": {
"command": "miyami-websearch-mcp"
}
}
}
`

That's it! Restart Claude Desktop and you're ready to search the web! 🎉

🚀 Quick Start

After adding to Claude Desktop config and restarting, try these prompts:

`
Search for the latest news about AI
`

`
Search for Python tutorials and summarize the top result
`

`
Fetch the content from https://example.com and summarize it
`

🛠️ Available Tools

$3

Search the web using multiple search engines with optional time-range filtering.

Parameters:
- query (required) - Your search query
- categories (optional) - general, news, images, videos, science
- language (optional) - Language code (default: en)
- page (optional) - Page number (default: 1)
- time_range (optional) - NEW! Filter by recency: day, week, month, year

Examples:
`
Search for "quantum computing breakthroughs" in news category
`

`
Search for AI news from the past 24 hours with time_range: day
`

`
Find recent Python tutorials from the past week with time_range: week
`

$3

Extract clean content from any webpage using Trafilatura (Firecrawl-quality extraction).

Parameters:
- url (required) - The webpage URL
- include_links (optional) - Include links (default: true)
- include_images (optional) - Include images (default: true)
- max_content_length (optional) - Max length in characters (default: 50000)
- format (optional) - Output format: text, markdown (default), html
- extraction_mode (optional) - Engine: trafilatura (default, best quality), readability (faster)
- stealth_mode (optional) - NEW! Anti-bot bypass: off, low, medium, high (default: off)
- auto_bypass (optional) - NEW! Auto-escalate stealth if bot protection detected (default: false)

Enhanced Features:
- 📝 Markdown output - Get structured markdown like Firecrawl
- 🎯 Rich metadata - Authors, dates, site names automatically extracted
- 📊 Extraction stats - Word count, content length, format info
- 🛡️ Stealth mode - Bypass Cloudflare, DataDome, Akamai, etc.

Example:
`
Fetch and summarize https://en.wikipedia.org/wiki/Artificial_intelligence in markdown format
`

$3

Search and automatically fetch full content from top results with Trafilatura-quality extraction.

Parameters:
- query (required) - Your search query
- num_results (optional) - How many results to fetch (1-5, default: 3)
- categories (optional) - Search categories
- time_range (optional) - Filter by recency: day, week, month, year
- format (optional) - Output format: text, markdown (default), html
- stealth_mode (optional) - NEW! Anti-bot bypass: off, low, medium, high (default: off)
- auto_bypass (optional) - NEW! Auto-escalate stealth if bot protection detected (default: false)

What it does:
- ✅ Searches for your query (with optional time filter)
- ✅ Gets top N results
- ✅ Automatically fetches full content (parallel)
- ✅ Uses Trafilatura for Firecrawl-quality extraction
- ✅ Returns both search snippets AND full webpage content
- ✅ FREE stealth mode for protected sites

Examples:
`
Research "climate change solutions" and give me detailed info from top 3 sources
`

`
Get recent AI breakthroughs from past 24 hours with full articles (time_range: day, num_results: 5)
`

`
Research recent web development tutorials from past week (time_range: week, format: markdown)
`

$3

Perform comprehensive parallel research across multiple topics at once with AI-powered reranking and compiled markdown reports.

Parameters:
- queries (required) - Comma-separated list of research queries (max 10)
- breadth (optional) - Results to fetch per query (1-5, default: 3)
- time_range (optional) - Filter by recency: day, week, month, year
- max_content_length (optional) - Max content per result (default: 30000)
- stealth_mode (optional) - Anti-bot bypass: off, low, medium, high (default: off)
- auto_bypass (optional) - Auto-escalate stealth if bot protection detected (default: false)

What it does:
- ✅ Process up to 10 queries in parallel for speed
- ✅ AI reranking for better relevance (always enabled)
- ✅ Auto-generates compiled markdown report
- ✅ Rich metadata extraction (author, date, source)
- ✅ Server-side caching (30 minutes)
- ✅ Aggregated statistics across all queries
- ✅ FREE stealth mode for protected sites

Examples:
`
Research "AI trends 2024,machine learning basics,ChatGPT use cases" with deep_research
`

`
Deep research on "React vs Vue,Next.js features,frontend trends" from past month
`

`
Comprehensive research: "climate solutions,renewable energy,carbon capture" with breadth: 5
`

$3

Depth-limited site crawler powered by Scrapy + Trafilatura. Returns structured pages with content, metadata, links, and word counts. Supports FREE stealth mode.

Parameters:
- start_url (required) - Starting URL to crawl
- max_pages (optional) - Max pages to crawl (1-200, default: 50)
- max_depth (optional) - Link depth (0-5, default: 2)
- format (optional) - Output format: text, markdown (default), html
- include_links (optional) - Include extracted links (default: true)
- include_images (optional) - Include image URLs (default: true)
- url_patterns (optional) - Comma-separated regex to include (e.g. /blog/,/docs/)
- exclude_patterns (optional) - Comma-separated regex to exclude
- stealth_mode (optional) - Anti-bot bypass: off, low, medium, high (default: off)
- obey_robots (optional) - Respect robots.txt (default: true; set false to bypass)

What it does:
- ✅ Depth-limited recursive crawling (Scrapy subprocess)
- ✅ Trafilatura extraction with metadata + word counts
- ✅ Include/exclude URL filtering
- ✅ FREE stealth mode with optional auto-escalation on server
- ✅ 15-minute crawl timeout and 30-minute cache

Examples:
`
Crawl docs site: start_url=https://docs.example.com max_depth=3 url_patterns=/api/,/guides/
`

`
Bypass robots on a small crawl: start_url=https://site.com max_pages=5 obey_robots=false stealth_mode=high
`

`
Filter sections: start_url=https://blog.example.com url_patterns=/2024/,/tech/ exclude_patterns=/archive/
`

$3

Fetch transcripts/captions from YouTube videos for LLM consumption. Supports multiple formats, language selection, translation, and time-range slicing.

Parameters:
- video (required) - YouTube video URL or 11-character video ID (supports all formats: full URL, youtu.be, embed, shorts)
- format (optional) - Output format: text (default), json (with timestamps), srt (subtitles)
- lang (optional) - Preferred language code (e.g., en, es, hi, fr). Default: auto
- translate (optional) - Translate transcript to target language code
- start (optional) - Start time in seconds for trimming
- end (optional) - End time in seconds for trimming
- list_langs (optional) - List available transcript languages instead of fetching (default: false)

What it does:
- ✅ Extract transcripts from any YouTube video with captions
- ✅ Multiple output formats (plain text, JSON with timestamps, SRT subtitles)
- ✅ Language selection for multilingual videos
- ✅ Translation to any supported language (via YouTube)
- ✅ Time-range slicing for specific segments
- ✅ List available transcript languages
- ✅ Stats: word count, segment count, duration
- ✅ 1-hour server-side caching

Examples:
`
Get transcript from YouTube video: video=dQw4w9WgXcQ format=text
`

`
Get transcript with timestamps: video=https://www.youtube.com/watch?v=dQw4w9WgXcQ format=json
`

`
Get Spanish transcript: video=dQw4w9WgXcQ lang=es
`

`
Translate to French: video=dQw4w9WgXcQ translate=fr
`

`
Get specific time range (60-120 seconds): video=dQw4w9WgXcQ start=60 end=120
`

`
List available languages: video=dQw4w9WgXcQ list_langs=true
`

💡 Usage Examples

$3


`
Use search_and_fetch to research "artificial general intelligence latest developments"
from the top 3 results and give me a comprehensive summary
`

$3

`
Search for AI breakthroughs from the past 24 hours using time_range: day
`

$3

`
Find Python tutorials from the past week using search with time_range: week
`

$3

`
Fetch this article in markdown format: https://example.com/article
`

$3

`
Use search_and_fetch to research "quantum computing" from the past week
with time_range: week and get full article content in markdown
`

$3

`
Search for "best restaurants in Tokyo" and show me the top 5 results
`

$3

`
1. Search for "Python web scraping libraries"
2. Fetch the documentation page in markdown format
3. Explain how to use it with examples
`

🔧 Configuration

No configuration needed! 🎉

This MCP server connects to a free public API automatically. Just add it to your Claude Desktop config and it works immediately.

If you're looking for advanced configuration options, there aren't any - we've kept it simple on purpose!

🐛 Troubleshooting

$3

1. Check your claude_desktop_config.json is valid JSON
2. Restart Claude Desktop completely (Quit and reopen)
3. Check Console.app (macOS) for error messages

$3

This is normal! The free tier API sleeps after inactivity. Subsequent requests are fast.

$3

The backend API is on Render free tier and may be waking up. Wait 60 seconds and retry.

$3

1. Ensure you have Node.js 18+ installed: node --version`
2. Try global install instead of npx
3. Check GitHub Issues

📡 API Backend

This MCP server connects to a free public API:
- URL: https://websearch.miyami.tech (hardcoded, no config needed)
- Cost: 100% Free - no API keys or signup required
- Privacy: No logging, no tracking, no data collection
- Engines: Google, DuckDuckGo, Bing, Brave, Wikipedia, Startpage
- Stealth Mode: FREE anti-bot bypass (Cloudflare, DataDome, Akamai, etc.)

🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

$3

Found a bug? Open an issue

📄 License

MIT License - see LICENSE file for details

🌟 Star History

If this tool helps you, please star the repo! ⭐

---

Made with ❤️ for the LLM community

Connect your AI to the internet in seconds, not hours.